Laboratory of Information Systems Design
Developing Optimal Intercloud Systems
Recent developments in cloud technologies enable easy, on-demand usage of computational resources via the network. We research techniques for controlling and using a global network of cloud systems (intercloud system).
National-scale intercloud and supercomputer cooperation
In Japan, many centers such as Hokkaido University have supercomputers that are connected with a high-speed 100Gbps network. This has many advantages, for example, computational simulations can produce output on the terabyte or petabyte scale. In applications such as optimizing aircraft design using simulations, it can be necessary to link or distribute the simulation among supercomputers. We develop intercloud systems that allow for data sharing, distributed processing and optimization on a national scale.
The image demonstrates the basic idea. A simulation is performed using a supercomputer, and the output stored in a distributed database using computer systems throughout the country. Cloud systems use this data to do optimization and visualization.
Optimal resource allocation and deployment in the Intercloud
There are now many cloud services, platforms and software available internationally. This is convenient, but users must choose the appropriate service or combination of services, develop their applications while considering the relative costs, conditions and compatibility of each service. This requires solving difficult optimization and constraint satisfaction problems.
In the ideal system to be developed, the user simply specifies the requirements of their application and a system chooses an appropriate combination of services automatically. We do research on resource-allocation and optimization systems necessary to realize this vision, while developing and improving the underlying algorithms.